3.1.

Plasma-Enhanced Chemical Vapor Deposition Grown SiO₂ and Si₃N₄ Caps

Figures 2(a) and 2(b) show the normalized 77 K PL spectra obtained from the PECVD ${\mathrm{SiO}}_2$ and ${\mathrm{Si}}_3{\mathrm{N}}_4$ capped SAQD samples at different annealing temperatures. Also, PL spectrum of the AG sample is included for reference. Notice in Fig. 2(a) that increasing the annealing temperature of ${\mathrm{SiO}}_2$ capped samples from 650°C to 725°C caused a gradual blue shift in the QD PL as compared to the AG SAQD sample. This implies IFVD facilitated Ga vacancies started to promote intermixing right from low-annealing temperatures.⁹ In addition, the PL intensity improvement (Table 1) and decrease in the PL linewidth [Fig. 2(c)] is noted, which is ascribed to the improved dot size homogeneity and reduction in the grown-in-defects due to low-temperature growth of QDs. From Fig. 2(c), a GS differential shift of $~145\mathrm{nm}$ (red arrow) was observed between the AG and 725°C annealed ${\mathrm{SiO}}_2$ sample, with associated linewidth reduction to $\sim 25\mathrm{nm}$ ($\sim 40\%$ of the AG sample, obtained via Gaussian fit). The PL peak intensity improvement was also found to improve by $>2$ times of the AG sample, as shown in Table 1. In general, the successive QD bandgap shift with increasing annealing temperature is significantly attributed to the effective sinking of Ga by ${\mathrm{SiO}}_2$ film and enhanced vacancy diffusion deeper into the sample due to high compressive strain developed at ${\mathrm{SiO}}_2/\mathrm{GaAs}$ interface and weakly to the thermal annealing effect. It is also worth mentioning that higher annealing temperatures reduced the intersublevel energy spacing of QD GS and ES emissions with observation of a single emission peak at 725°C, possibly due to masking of the ES emission by the dominant GS peak or by the detection limit of our PL system. However, in the case of ${\mathrm{Si}}_3{\mathrm{N}}_4$ capped sample at 725°C, as illustrated in Fig. 2(b), a clear two-peak QD PL is visible. This implies a comparatively slow interdiffusion rate offered by ${\mathrm{Si}}_3{\mathrm{N}}_4$ film and is further substantiated by and an overlapping PL spectrum of 650°C annealing temperature sample with the AG sample. In other words, ${\mathrm{Si}}_3{\mathrm{N}}_4$ layer works as a control cap to partially inhibit diffusion of vacancies in the SAQD structure at this particular temperature. We postulated small group-V out-diffusion from the ${\mathrm{Si}}_3{\mathrm{N}}_4$ capped sample might be responsible for suppression of inherent thermal shift,² and negligible IFVD effect possibly due to inefficient Ga vacancies generation (insolubility of Ga or As atoms in ${\mathrm{Si}}_3{\mathrm{N}}_4$) and diffusion into the SAQD sample (smaller compressive stress at ${\mathrm{Si}}_3{\mathrm{N}}_4/\mathrm{GaAs}$ interface)²⁵ at this low temperature. Nevertheless, at 675°C and beyond, the PL spectra blue shifts in conjunction with narrow GS and ES PL linewidths, and reduction in their energy spacing, at high annealing temperatures, as depicted in Fig. 2(c). This is indicative of enhanced interdiffusion mediated through competing thermal annealing and IFVD process. Note that our observations are in agreement with most reported results^2,27,32 on similar In(Ga)As/GaAs QD material system. These observations are in contrast to the work of Wang et al.,²³ which reported a high degree of intermixing from ${\mathrm{Si}}_3{\mathrm{N}}_4$ cap compared to ${\mathrm{SiO}}_2$ on InAs/InGaAlAs/InP QD-in-well material system, and attributed to dominant In outdiffusion with respect to Ga. However, we believe Ga outdiffusion to be dominant in our InAs/GaAs SAQD sample. Referring to Figs. 2(a) and 2(b), it is worth mentioning that selectively intermixing SAQD sample with ${\mathrm{Si}}_3{\mathrm{N}}_4$ and ${\mathrm{SiO}}_2$ film at 650°C may result in ultrabroad simultaneous and comparable PL emission from QD GS and ES, with $-3\mathrm{dB}$ bandwidth spanning $\sim 250\mathrm{nm}$. Devices exhibiting such broad emissions are highly attractive for applications in optical communication, such as broadband lasers, detectors, modulators, etc.,^22,28 discussed later in this work. In addition, a blue shifted and broad emission with peak PL intensity at ES rather than GS with linewidth of $\sim 78\mathrm{nm}$ is observed with ${\mathrm{Si}}_3{\mathrm{N}}_4$ capped SAQD sample at an intermediate temperature of 675°C (GS emission shoulder is still visible at $\sim 1200\mathrm{nm}$). In fact, analogous behavior is noticed in the PL results of other capping materials (${\mathrm{HfO}}_2$, ${\mathrm{Al}}_2{\mathrm{O}}_3$, ZnO, and ${\mathrm{TiO}}_2$ caps) at low and intermediate annealing temperatures. We attribute this observation to an increased interface fluctuation between SAQDs and the surrounding matrix at low and intermediate temperatures, and possible defect annealing near the interface.^2,33 This improved with increasing temperature as a result of higher In–Ga intermixing. The normalized integrated PL intensity and the GS peak PL intensity of ${\mathrm{Si}}_3{\mathrm{N}}_4$ and ${\mathrm{SiO}}_2$ capping layers at different annealing temperatures are summarized in Table 1. The results suggest that the optical quality of the SAQDs is maintained at low and intermediate temperatures via possible reduction of the defect density and the nonradiative recombination centers. At relatively high annealing temperature, enhanced In–Ga intermixing resulted in improvement of the QD size distribution and the structural quality.^20,23

Fig. 2

Normalized 77 K photoluminescence (PL) spectra obtained from the AG and annealed samples of InAs/GaAs self-assembled InAs/GaAs quantum dots (SAQDs) at different annealing temperatures for plasma-enhanced chemical vapor deposition (PECVD) deposited: (a) ${\mathrm{SiO}}_2$, and (b) ${\mathrm{Si}}_3{\mathrm{N}}_4$ dielectric capping layers. Summary of the corresponding change in GS peak emission and linewidth as a function of annealing temperatures. The solid horizontal lines near both the vertical axes of (c) correspond to the as-grown values.

Table 1

Summary of the ground-state (GS) peak photoluminescence (PL) intensity and the total integrated PL intensity (shown in the parenthesis), normalized to the as-grown sample’s GS peak PL intensity and the total integrated PL intensity, respectively, at different annealing temperatures and for different capping materials.

3.2.

Pulsed Laser Deposition Grown HfO₂ and SrTiO₃ Caps

Figures 3(a) and 3(b) show the PL spectra of the samples capped with 100 nm PLD ${\mathrm{HfO}}_2$ and ${\mathrm{SrTiO}}_3$ caps at different annealing temperatures. The corresponding changes in the peak GS emission wavelength and PL linewidth are summarized in Fig. 3(c). The ${\mathrm{HfO}}_2$ capped sample shows inhibition to intermixing at 650°C with no differential shift in the GS peak emission. The ES PL intensity is found to increase in this case, similar to the ${\mathrm{Si}}_3{\mathrm{N}}_4$ capped samples at intermediate temperature. Apart from possible defect annealing near the QD/barrier interface,² we attribute this observation to the minor outdiffusion of Ga from smaller size QDs in particular due to comparatively larger surface to volume ratio, than larger size QDs at low annealing temperatures.²³ As the temperature is increased to 675°C and to 700°C, a high degree of intermixing with a differential GS PL peak shift (compared to AG sample) of $\sim 175\mathrm{nm}$ was noted. Moreover, observation of $\sim 4.0$ (2.5) times increase (decrease) in the GS peak PL intensity (linewidth) compared to the AG sample further substantiate this material being a highly efficient intermixing cap, as summarized in Table 1. Note that IFVD effect saturated at 700°C with no further blue shifting of the GS photoluminescence peak beyond this annealing temperature. Moreover, a single peak emission at 725°C possibly indicates merging of GS and ES emissions. The PL linewidth at this annealing temperature was found to be similar in value ($\sim 18\mathrm{nm}$) compared to the 700°C annealed sample except with an increase in the peak PL intensity ($\sim 50\%$) and comparable integrated PL intensity. This might be due to complete inhabitation of intrinsic defects, possible oxygen vacancies, in the 100-nm thick ${\mathrm{HfO}}_2$ film by Ga outdiffusion.³⁴ In addition, the alteration of strain at the ${\mathrm{HfO}}_2/\mathrm{GaAs}$ interface (possibly from thermal matching to tensile strained GaAs surface region) at elevated annealing temperatures could also play a role in suppressing the interdiffusion process.^27,32 A selection of a thick ${\mathrm{HfO}}_2$ capping layer would probably enable observation of intermixing effect beyond 700°C. In general, annealing SAQD samples with ${\mathrm{SiO}}_2$ and ${\mathrm{HfO}}_2$ caps enhanced the QD luminescence with spectral narrowing and has been attributed to the improvement in the QD inhomogeneity. On the contrary, dissolving of QDs into QD-QW metamorphic structure (two-dimensional-like system) at high temperatures is also possibly due to strong lateral In–Ga interdiffusion in the 3-D QD structures, and might lead to improved material quality, as indicated in Refs. 28 and 35. In either case, an emission window of 1030–1060 nm at 77 K translates to $\sim 1100─1160\mathrm{nm}$ at room temperature. Hence, employing high temperature ($>700°\mathrm{C}$) annealed ${\mathrm{SiO}}_2$ or ${\mathrm{HfO}}_2$ capped SAQD lasers with superior characteristics would enable realization of frequency-doubled green–orange–yellow band lasers.¹⁵ Alternatively, SAQDs deposited with the ${\mathrm{SrTiO}}_3$ film are found to be a highly effective cap for intermixing suppression, as depicted in Fig. 3(b). A remarkable inhibition to the thermal shift and Ga vacancy diffusion up to 700°C is demonstrated with a mere $\sim 16\mathrm{nm}$ GS differential shift compared to the AG sample. Moreover, the PL intensity and the PL linewidth are found to be analogous to the AG sample, thus representative of preserving the optical quality of the annealed samples, as depicted in Table 1. We postulate that a high degree of tensile stress created at the ${\mathrm{SrTiO}}_3/\mathrm{GaAs}$ interface impede the down diffusion of Ga vacancies and thus inhibit group-III intermixing between QDs and surrounding barrier layers. A large thermal expansion coefficient of the ${\mathrm{SrTiO}}_3$ cap, which is $\sim 22\%$ larger than the typical ${\mathrm{TiO}}_2$ cap, further upholds our postulation.²⁷ On the other hand, ${\mathrm{SrTiO}}_3$ may also cause reduction of Ga vacancies generation possibly due to various factors during annealing such as layer quality, diffusion of inherent defects, and the metallurgical reaction between GaAs and ${\mathrm{SrTiO}}_3$ films.³² Note that, at 725°C, a PL blue shift accompanied by a broad emission with peak at ES rather than GS (shoulder at longer wavelength region) is observed, indicating that the limit for intermixing inhibition is up to 700°C. We postulate that the emission peaks observed at 1125 and 1050 nm is from the GS and ES of highly interdiffused QDs with small sizes, and the long wavelength shoulder to the GS emission of least interdiffused QDs with larger size where the interdiffusion is minimal. We also believe that the shift in the GS peak emission wavelength at 725°C is also controlled by the intrinsic thermal annealing induced disordering effect.²

Fig. 3

Normalized 77 K PL spectra obtained from the as-grown and annealed samples of InAs/GaAs SAQDs at different annealing temperatures for pulsed laser deposition (PLD) deposited: (a) ${\mathrm{HfO}}_2$, and (b) ${\mathrm{SrTiO}}_3$ dielectric capping layers. Summary of the corresponding change in GS peak emission and linewidth as a function of annealing temperatures. The solid horizontal lines near both the vertical axes of (c) correspond to the as-grown values.

3.3.

Atomic Layer Deposition Grown Al₂O₃, ZnO, and TiO₂ Caps

Low temperature PL spectra of ALD deposited ${\mathrm{Al}}_2{\mathrm{O}}_3$, ZnO, and ${\mathrm{TiO}}_2$ capped SAQD samples at annealing temperatures from 650°C to 725°C are plotted in Figs. 4(a)–4(c), respectively. For all capping materials, we observed inhibition of intermixing at a low-annealing temperature of 650°C, similar to the ${\mathrm{Si}}_3{\mathrm{N}}_4$, ${\mathrm{HfO}}_2$, and ${\mathrm{SrTiO}}_3$ caps, and the same reason holds for these ALD grown dielectric films. As the annealing temperature is increased beyond 650°C, a progressive blue shift in the GS peak PL emission is observed in all three (${\mathrm{Al}}_2{\mathrm{O}}_3$, ZnO, and ${\mathrm{TiO}}_2$) samples, as shown in Fig. 4(d). In all the cases, an improvement in the material quality is noted and summarized in Table 1. However, GS and ES peaks did not merge in all the three capped samples even at the highest annealing temperature of 725°C. Distinct peak emissions from both the transition states are clearly observed in ${\mathrm{Al}}_2{\mathrm{O}}_3$ and ZnO capped samples, whereas the ${\mathrm{TiO}}_2$ capped sample showed bimodal peak behavior, a result of reduction in the intersublevel energy spacing of QD GS and ES emissions. Nonetheless, the ${\mathrm{TiO}}_2$ film showed the highest degree of intermixing when compared to the other two ALD films with a GS peak blue shift of $\sim 165\mathrm{nm}$ and similar PL linewidth value compared to the AG data, as elaborated in Fig. 4(d). This observation is most likely due to our ALD technique that might alter the film porosity and thus promote intermixing,²⁵ as compared to the literature^27,28 where e-beam evaporated ${\mathrm{TiO}}_2$ film was shown to inhibit intermixing. On the other hand, 675°C and 700°C annealed ZnO capped samples showed a small GS wavelength blue shift of $\sim 32\mathrm{nm}$ compared to the AG sample indicative of partial inhibition of the IFVD process. However, the GS PL linewidth broaden at this temperature by $\sim 40\%$ (compared to the AG sample) and is accompanied by increased luminescence from the ES of QDs. This suggests increased QD size and composition dispersion as observed in other PECVD and ALD samples. In general, at low-annealing temperatures (650°C and 675°C) most of the capping layers showed increased GS PL linewidth, which has been attributed to the interface fluctuation between SAQDs and the surrounding matrix, thus affecting the QD transition states. In addition, an inhomogeneous rate of In–Ga diffusion in different QD sizes given 3-D and complex intermixing process could also be ascribed for this observation.² In the case of the ${\mathrm{Al}}_2{\mathrm{O}}_3$ capped sample, typical blue shifting of the GS peak wavelength is observed with progressive narrowing and enhancement of the PL linewidth and intensity, respectively, with a maximum peak shift of $\sim 125\mathrm{nm}$ at 725°C compared to the reference AG data. In general, this layer behaves similar to the other IFVD promoting caps (${\mathrm{SiO}}_2$, ${\mathrm{HfO}}_2$, and ${\mathrm{Al}}_2{\mathrm{O}}_3$) and retains (improves) the optical quality of the material after low (high) temperature annealing, as illustrated in Table 1. A remarkable observation that is worth mentioning about this film is the broadened emission with equal intensities from both GS and ES transitions of QDs annealed at low temperature (650°C), indicating increased compositional fluctuation at the interface between the QD and the surrounding matrix as compared to ${\mathrm{TiO}}_2$ and ZnO capping, leading to a dispersive QD potential profiles, in particular, affecting the smaller dots with higher intermixing rate as discussed earlier.³⁶ An ultrabroad PL linewidth of $\sim 165\mathrm{nm}$ from this single capped low-temperature intermixed SAQD structure, centered at $\sim 1140\mathrm{nm}$ is again highly attractive for biomedical imaging, in the low-coherence interferometry system such as optical coherence tomography.^13,22

Fig. 4

Normalized 77 K PL spectra obtained from the as-grown and annealed samples of the InAs/GaAs QDs at different annealing temperatures for atomic layer deposition (ALD) deposited: (a) ${\mathrm{Al}}_2{\mathrm{O}}_3$, (b) ZnO, and (c) ${\mathrm{TiO}}_2$. Summary of the corresponding change in GS peak emission and linewidth as a function of annealing temperatures. The solid horizontal lines near both the vertical axes of (d) correspond to the as-grown values.

3.4.

Comparative Analysis

To compare the performance of different capping layers, we extracted two parameters that characterize their degree of intermixing, namely the intermixing rate ($\delta \mathrm{E}$) and the critical temperature (${\mathrm{T}}_{\mathrm{C}}$). Figures 5(a)–5(c) and 6(a)–6(c) show the GS peak energy shift, and the energy separation between GS and ES ($\mathrm{\Delta }\mathrm{E}$), extracted from the 77 K PL spectra, as a function of annealing temperature, and corresponding to PECVD, PLD, and ALD deposited capping materials, respectively. The results are summarized in Table 2. For simplicity in analysis, we selected the region that showed linear interdiffusion (promotion by all the samples except ${\mathrm{SrTiO}}_3$ and ZnO which showed inhibition of interdiffusion) and $\mathrm{\Delta }\mathrm{E}$ behavior, for estimating $\delta \mathrm{E}$ and ${\mathrm{T}}_{\mathrm{C}}$,^9,24 respectively, with $\pm 20°\mathrm{C}$ margin of error in the latter case ($\mathrm{\Delta }\mathrm{E}$ values are extracted from GS and ES Gaussian fittings). A reasonably good linear fitting is obtained in both the parameter extraction cases, as depicted in Figs. 5 and 6, with close agreement of ${\mathrm{T}}_{\mathrm{C}}$ value obtained from linear interpolation and the experiments. For instance, possible merging of QD GS and ES emissions is observed in ${\mathrm{HfO}}_2$ and with some uncertainty in ${\mathrm{SiO}}_2$ and ${\mathrm{Al}}_2{\mathrm{O}}_3$ capped samples, at 725°C; meanwhile, the corresponding extracted ${\mathrm{T}}_{\mathrm{C}}$ values are 725°C, 770°C, and 788°C, respectively. The rate of intermixing is found to be highest for the ${\mathrm{HfO}}_2$ film with $\delta \mathrm{E}=3.7\mathrm{meV}/°\mathrm{C}$ and ${\mathrm{T}}_{\mathrm{C}}=725°\mathrm{C}$. Compared to ${\mathrm{SiO}}_2$ film, an increase in $\delta \mathrm{E}$ by a factor of more than 3 ($1.05\mathrm{meV}/°\mathrm{C}$ for ${\mathrm{SiO}}_2$) with similar ${\mathrm{T}}_{\mathrm{C}}$ values is noted. This suggests that PLD ${\mathrm{HfO}}_2$ cap has better capability to promote In–Ga interdiffusion between barrier layers and QDs compared to PECVD ${\mathrm{SiO}}_2$ cap. A similar increase in $\delta \mathrm{E}$ value (by a factor of 2.5–4.5) is also observed on comparing ${\mathrm{HfO}}_2$ capped sample with other IFVD promoting caps, i.e., ALD deposited ${\mathrm{TiO}}_2$ ($1.46\mathrm{meV}/°\mathrm{C}$) and ${\mathrm{Al}}_2{\mathrm{O}}_3$ ($0.77\mathrm{meV}/°\mathrm{C}$) films; the latter also showing a relatively higher degree of intermixing with larger (smaller) $\delta \mathrm{E}$ (${\mathrm{T}}_{\mathrm{C}}$) values. Most likely this observation might be attributed to the quality of the capping layer with possible increase in porosity for the PLD film compared to ALD caps, which changes the interdiffusion rate.²⁵ In contrast, the extracted slope of the linear fit for PLD ${\mathrm{SrTiO}}_3$ capped sample data (within the linear region) with value $\delta \mathrm{E}=0.31\mathrm{meV}/°\mathrm{C}$ and covering higher annealing temperatures signifies strong IFVD inhibition compared to ZnO ($0.22\mathrm{meV}/°\mathrm{C}$ but obtained with smaller annealing temperature range) and ${\mathrm{Si}}_3{\mathrm{N}}_4$ ($1.00\mathrm{meV}/°\mathrm{C}$) caps with similar predicted ${\mathrm{T}}_{\mathrm{C}}$ value of 856°C (compared to 842°C and 822°C, respectively). Thus, ${\mathrm{SrTiO}}_3$ is more attractive due to its ability to inhibit intermixing within a larger and higher range of annealing temperatures. Note that ${\mathrm{Si}}_3{\mathrm{N}}_4$ dielectric layer, in our case, behaved more as an intermixing-induced film rather than an inhibiting cap at high-annealing temperatures, and is found to be in good agreement with other reports.^2,10,23 The possible reason for this efficient control of QD emission during IFVD process by ${\mathrm{SrTiO}}_3$ is the large tensile stress created at the surface, more than a factor of 2, compared to the ZnO and ${\mathrm{Si}}_3{\mathrm{N}}_4$ caps. Therefore, this possibly resulted in more effective defect agglomeration and cluster formation.² The $\delta \mathrm{E}$ value of ${\mathrm{SrTiO}}_3$ is found to be one order of magnitude smaller than ${\mathrm{HfO}}_2$ cap, providing a differential bandgap shift of $\sim 175\mathrm{nm}$ between them. We also observed that in spite of the $\delta \mathrm{E}$ value being similar for ${\mathrm{Al}}_2{\mathrm{O}}_3$, ${\mathrm{TiO}}_2$, ${\mathrm{Si}}_3{\mathrm{N}}_4$, and ${\mathrm{SiO}}_2$ ($\sim 1.0\pm 0.3\mathrm{meV}/°\mathrm{C}$) capped samples, the extracted ${\mathrm{T}}_{\mathrm{C}}$ values of ${\mathrm{TiO}}_2$ and ${\mathrm{Si}}_3{\mathrm{N}}_4$ films are found to be larger than the other capping layers. We postulate that the increased outdiffusion of Ga from small dots (dominating the ES emission) as compared to the larger dots (dominating the GS emission) might be responsible for this observation.

Fig. 5

Linear fitting of the GS peak energy shift as a function of annealing temperatures: (a) PECVD grown ${\mathrm{SiO}}_2$ and ${\mathrm{Si}}_3{\mathrm{N}}_4$, (b) PLD grown ${\mathrm{HfO}}_2$ and ${\mathrm{SrTiO}}_3$, and (c) ALD grown ${\mathrm{Al}}_2{\mathrm{O}}_3$, ${\mathrm{TiO}}_2$, and ZnO capped InAs/GaAs SAQD structure, respectively. The slope of the linear fit provides the rate of interdiffusion $\delta \mathrm{E}$. The solid horizontal line near the vertical axis corresponds to the as-grown GS peak energy.

Fig. 6

Linear fitting of the GS and ES peak energy separation as a function of annealing temperatures: (a) PECVD grown ${\mathrm{SiO}}_2$ and ${\mathrm{Si}}_3{\mathrm{N}}_4$, (b) PLD grown ${\mathrm{HfO}}_2$ and ${\mathrm{SrTiO}}_3$, and (c) ALD grown ${\mathrm{Al}}_2{\mathrm{O}}_3$, ${\mathrm{TiO}}_2$, and ZnO capped InAs/GaAs SAQD structure, respectively. The critical temperature ${\mathrm{T}}_{\mathrm{C}}$ is calculated by subsequent linear extrapolation. The solid horizontal line near the vertical axis corresponds to the AG $\mathrm{\Delta }\mathrm{E}$ value.

Table 2

Summary of the extracted δE and TC values of different IFVD processed InAs/GaAs SAQD capping materials.

3.5.

Applications

Our analysis of characterizing the degree of intermixing via $\delta \mathrm{E}$ and ${\mathrm{T}}_{\mathrm{C}}$ qualitatively provided a series of data to exploit seven different capping layers as interdiffusion suppressors or promoters for various applications. In the following, we highlight three potential applications employing either one or a combination of different caps: